Transparent Process

How your selfie becomes lipstick recommendations

This page explains the method in plain language so every recommendation feels clear and grounded.

Matching pipeline

  1. Landmark detection: We locate key points around eyes, nose, and facial structure using face-api.js.
  2. Cheek sampling: We sample both cheek regions, then use median RGB values to reduce noise from highlights and shadow.
  3. Tone + undertone estimation: RGB/HSL values are mapped into skin tone depth and warm, cool, or neutral undertone.
  4. Shade ranking: We rank lipstick shades by compatibility and return three clear choices.

Why this method feels consistent

  • The process is deterministic and repeatable with similar photos.
  • Cheek-based sampling is used because it is less biased than lips or forehead alone.
  • You can inspect your sampled color, detected tone, and detected undertone before buying.

How to improve accuracy

  • Use soft daylight or evenly lit indoor light.
  • Keep your face centered, straight, and unobstructed.
  • Avoid intense beauty filters or overexposed flash photos.

FAQ

What if my result changes between photos?

Lighting and camera settings can shift sampled color, which affects shade ranking.

Can I use this without makeup?

Yes. A makeup-free selfie usually gives the most neutral skin-tone sample.

Where do the shade links go?

Each recommendation opens a product listing so users can compare and purchase quickly.

Try the matcher now

Go back to the main tool and test your selfie in under a minute.